AI visibility audit for car dealerships that reads every answer, then fixes what the engines read
Ask ChatGPT which dealerships in your city to avoid. If your store is in that answer, a tool that counts mentions scores it as a win. I ask at least 12 buyer questions, the avoid question every time, in two separate passes, signed out, and I read every answer your store appears in. Then I fix the sources the engines read about your store, and run the same questions again every month.
A score out of 100 tells you how often you were named. It doesn't change a word of what the engines read.
Request your market's indexThe Index: your store and the stores around you, against the questions buyers ask, signed out and dated. Request it and your store's first answers are pulled before a 15-minute call.
One capture is a snapshot. Here is what four looked like.
Hickory, North Carolina. Perplexity, counted by exact store name. The same 15 questions, four runs across 22 days. Seven stores went unnamed in July, seven on 8 August, seven the morning of 14 August, and nine that afternoon. Even a plain name count shows the answers moving between morning and afternoon.
Four stores were never named once in any of the 60 answers. Three of them are franchise stores: a Hyundai store, a Mitsubishi store, a Volkswagen store.
A report built on the morning run and a report built on the afternoon run would disagree about at least two stores. That's why every question in this audit runs at least twice, and why a count only tells you where to start reading.
What a car dealership AI visibility audit has to get right
It's fair to ask any audit how it was built. A mention count goes wrong in four places.
A single capture.
The order of names moves from one run to the next. Every question runs in at least two passes, so the record shows what held and what moved.
A warning scored as a win.
"Which car dealerships in your city should I avoid?" is asked on every audit. On one dealer group, that one question reversed our read on two stores we had scored as healthy. The whole avoid answer gets read, and your group's name gets checked under every spelling, because an answer can warn buyers off the group without ever naming the store.
The wrong business.
Every store is matched to its street address before a single question is asked. A same-name store in another state, a sister store in your own group, and a store listed under the neighboring city all look like real data to a counter. A loose name matcher once credited a business in 23 of 24 answers because those answers contained two common words from its name. Every count here runs with invented store names as a control, and they have to come back zero.
A cut-off answer.
An answer that comes back truncated, timed out, or with no answer in it looks exactly like your store being left out. Those get run again. They never get scored.
Then every mention is read by hand, in context. On one run, a keyword classifier caught 1 of 14 real demotions: answers that named a store and then steered the buyer somewhere else.
How the audit runs
The roster. The stores in your market, pulled from Google Maps under several searches, deduplicated, ranked by review count, matched to their addresses, and filtered to your city.
The questions. At least 12 per market. They cover buying, service, best-of, used, your brand, leasing, recalls, the most trustworthy store, and which dealerships to avoid.
The conditions. On ChatGPT, signed out, with a fresh browser for every question. Any answer captured while signed in is thrown out. When more than one engine is asked, they run one at a time, never side by side.
Same depth everywhere. When markets get compared, every market gets the same number of questions and the same number of passes.
Three answers a buyer can get about your store, and a fourth.
| State | What it means |
|---|---|
| Not named | The engine answered the question and your store wasn't in it. |
| Named but not chosen | You were in the list. The buyer was pointed somewhere else. |
| Recommended | The engine picked you, and said why. |
| On the avoid list | A separate flag. A mention counter scores it as a win. |
What the engine prints beside your store. The star rating it shows a buyer, and which listing that rating came from. On one store, ChatGPT printed a sister listing's rating and a service listing's lower one, and never printed the store's own, which was the highest of the three.
The sources it cites. Which sites each engine is reading about your store. That's the lever, and it's usually not Google.
What you receive
No passwords, no DMS access, no CRM access. The index and the monthly need no login from your side. In AI Source Repair, profiles get claimed under your login, so what gets corrected stays yours.
Before we talk. Request your market's index and within one business day you get an email from hello@storefrontaudit.com with times for a 15-minute call. Your store's first answers are pulled before the call. That's a first look. The full battery runs in AI Source Repair and every month on the monthly.
AI Source Repair, $797 one time. Your market's buyer questions run clean-state on ChatGPT, Google AI and Perplexity, and every source feeding each answer about your store gets traced. You get the full source map. The third-party profile feeding a wrong number gets claimed under your login, not ours. Corrections go to the aggregators mislabeling your store, every change dated, in a record you keep. At day 30 the identical capture runs again, and you get a before-and-after page and a review call. Everything it claims, corrects and documents is yours, whether or not you ever spend another dollar with us.
The monthly. The same battery runs on your market every month, signed out, dated and verbatim, so month over month is a real comparison and not a fresh opinion. You get what moved against last month's capture. Every month carries its own intervention list off that delta, and the work that month comes off the list. Month one clears the wreckage. The months after are when the position gets built, because the engines keep re-reading their sources and your competitors keep moving. Two AI vendor pitches a month get read against a fixed written scorecard and come back to you inside five business days. Forty-five minutes a month to walk it, plus one page you can carry upstairs and defend without me in the room.
Pricing, in writing
Groups are priced on the call, because the market count and the call change.
No ranking guarantees, no promised units, gross or lead counts. The engines re-read their sources constantly and answers move on their own. What is guaranteed is that it gets measured the same way every month and you see the delta in writing.
Who reads the answers
Adrian Marin spent eighteen years in car sales, from the floor to the general manager's office. The desk, the floor, the factory calls. The measurement is his, and so is the phone call.
Questions dealers ask
What is an AI visibility audit for a car dealership?
It's a dated record of what AI engines tell a car buyer about your store. At least 12 buyer questions for your market, including which dealerships to avoid, asked in two passes, signed out. Every answer is read by hand, and your store is marked not named, named but not chosen, or recommended, with a separate flag when it lands on an avoid list. The record also shows the rating each engine prints beside your store and the sources it cites.
How is this different from an AI visibility score out of 100?
A score that counts mentions adds a warning to your total the same way it adds a recommendation. This audit keeps them apart. Each answer puts your store in one state (not named, named but not chosen, or recommended), avoid-list mentions are flagged on their own, and you see the answer itself, word for word.
How many questions, and how many times is each one asked?
At least 12 per market, and every one runs in at least two passes, because the order of names moves between runs. Each question gets a fresh, signed-out browser. An answer that comes back cut off or empty is run again instead of scored.
What should I ask anyone selling me an AI visibility audit?
Which engines they tested, which questions, in which market, on what dates, how many times each question was repeated, whether they were signed out, which sources the engines cited, whether every competing store was measured under the same conditions, and how an avoid-list mention is scored. This page answers every one of those before you request anything.
Do you need access to our website, DMS or CRM?
No. No passwords, no DMS access, no CRM access. The audit asks the engines the same questions your buyers ask, from the outside. When a source gets corrected in AI Source Repair, the profile is claimed under your login, not ours.
What does it cost, and do you guarantee results?
AI Source Repair is $797 one time, with no subscription attached. The monthly is $2,500 per store for franchise stores and $1,500 for a single independent rooftop, with no contract and no term. Groups are priced on the call. There are no ranking guarantees and no promised units, gross or lead counts, because the engines re-read their sources constantly and answers move on their own. What is guaranteed is that it gets measured the same way every month and you see the delta in writing.
Your store, measured the same way.
Request your market's indexWithin one business day you'll get an email from hello@storefrontaudit.com with times for a 15-minute call, and your store's first answers will be pulled before we talk.